@inproceedings{da9870a9f84349be9a857aba8419cdb8,
title = "Eye-Tracking-Based Image Annotation Behavior Analysis",
abstract = "The rapid development of artificial intelligence has significantly increased the demand for high-quality data annotation, which is crucial for optimizing models and enabling practical applications. However, manual annotation is flexible but often inefficient and costly. Additionally, insufficient quality control can lead to inconsistent annotations, which hinder AI model performance. Eye-tracking research, which provides valuable insights into shifts in attention, forms a foundation for understanding user attention patterns. GazeLabel is introduced as a tool integrating eye-tracking data with data annotation to evaluate annotation quality. It analyzes eye-tracking data from annotators using metrics such as first gaze duration, regression count, gaze-saccade ratio, Intersection over Union (IoU), and Consecutive Images Gaze Synchronization (CIGS). The system also provides both individual and group visualizations of eye-tracking data, aiding users in better understanding annotator behavior and the quality of annotations.",
keywords = "data annotation, evaluation of annotation quality, eye-tracking, visualization",
author = "Zhenqin Chen and Chaoquan Luo and Sentao Liu and Zhuo Yang and Ming Li",
note = "Publisher Copyright: {\textcopyright} 2025 SPIE.; 17th International Conference on Digital Image Processing, ICDIP 2025 ; Conference date: 25-04-2025 Through 27-04-2025",
year = "2025",
month = jul,
day = "22",
doi = "10.1117/12.3073037",
language = "English",
series = "Proceedings of SPIE - The International Society for Optical Engineering",
publisher = "SPIE",
editor = "Ting-Chung Poon and Xudong Jiang and Zhaohui Wang and Jindong Tian",
booktitle = "Seventeenth International Conference on Digital Image Processing, ICDIP 2025",
address = "United States",
}